An MVP is not a bad version of your product. It is the smallest thing that tests whether people will pay for a solution to their problem. Most founders get this wrong — they spend months coding a "minimum" product that is neither minimum nor viable.
The Concierge-to-Code Ladder fixes this by starting at the bottom: deliver the value by hand, prove people will pay, then automate one step at a time.
The three types of MVP
Type 1: Concierge MVP
You deliver the service entirely by hand. No code, no automation, no product. The customer gets the outcome they want; you do the work behind the scenes manually.
Example: Instead of building an AI meal-planning app, you email five customers a personalized weekly meal plan that you created in a spreadsheet. If they pay $20/week for this, you have validated demand. If they don't, no amount of code would have helped.
When to use: You have not yet confirmed that someone will pay for this outcome.
Type 2: Wizard of Oz MVP
It looks automated to the customer, but you run the process manually behind the scenes. The customer interacts with what appears to be a product; you are the product.
Example: You build a simple form where customers submit their dietary preferences. They receive a "generated" meal plan 24 hours later. In reality, you create each plan manually. To the customer, it feels like software.
When to use: You have confirmed demand (Concierge worked) and want to test the user experience before building the real product.
Type 3: Automated MVP
Real code, real product — but still the absolute minimum feature set. One workflow, one use case, one screen if possible. Everything that is not essential to the core job is cut.
When to use: Both Concierge and Wizard of Oz have validated demand and user experience. Now you need to remove yourself from the delivery loop.
The decision tree
Start here and move down:
- Have five people said they would pay? No → Go back to validation. Yes → Concierge MVP.
- Have five people actually paid for the manual version? No → Keep doing Concierge. Yes → Wizard of Oz MVP.
- Is the manual delivery costing you more than 20 hours/week? No → Keep Wizard of Oz. Yes → Automated MVP.
Most founders skip to step 3 on day one. That is the most expensive mistake in startups.
The 7-day MVP sprint
This sprint is designed for the Concierge level — the fastest path from idea to paid customer.
Day 1: Write the offer page
One page. One problem. One outcome. One price. One call to action. Use Carrd, Framer, or a simple Google Doc. The page must answer: what problem does this solve, who is it for, what do you get, and how much does it cost? No features. No roadmap. Just the outcome.
Day 2: Send it to 20 people
Not strangers on Twitter. Twenty people who you already know have the problem, or who are in a community where the problem is discussed. Use the Mom Test framework to avoid leading questions. Send a direct message, not a broadcast.
Days 3–5: Deliver manually
Anyone who says yes gets the service delivered by hand. This is where you learn what the product actually needs to do — not from a product spec, but from doing the work. Take notes on every step: what was easy, what was painful, what surprised you, what the customer valued most.
Day 6: Collect feedback
Ask three questions: What was most valuable? What was missing? Would you pay again next month? The answers tell you whether to continue, adjust, or kill.
Day 7: Decide
Three outcomes are possible:
- Kill. Nobody paid, or the feedback is that the problem is not painful enough. Read when to kill an idea.
- Iterate. Some people paid but the offer needs adjustment — different scope, different price, different audience.
- Build. Multiple people paid and want to continue. Now — and only now — start building the Wizard of Oz or Automated version.
What an MVP is not
- Not a prototype. A prototype demonstrates feasibility. An MVP demonstrates demand.
- Not a feature list. If your MVP has more than one core workflow, it is not minimum.
- Not free. A free MVP does not test willingness to pay. Charge from day one, even if the price is low.
- Not perfect. Rough edges are fine. Broken core value is not.
Tools for non-technical founders
You do not need to code to run an MVP:
- Offer page: Carrd ($19/year), Framer (free tier), or Google Docs (free)
- Payments: Stripe, Lemon Squeezy, or Gumroad
- Delivery: Notion (shared workspace), Loom (video walkthroughs), Email
- Automation (later): Make.com, Zapier, or n8n
- AI-assisted building (later): Cursor, Bolt, Lovable for when you need code
The rule that saves months
Do not write a line of code until a stranger has paid you for the manual version. If the manual version is not worth paying for, the automated version will not be either. The Concierge-to-Code Ladder exists because the most expensive thing in startups is building the wrong product — and the cheapest test is doing the job yourself first.
MoatKit's "Launch Your Startup" pathway walks founders through this exact sequence — from validation to first proof — in 14 lessons over three weeks. See the curriculum.